CELLULAR NETWORK
INFRASTRUCTURE
The Future of Fog Monitoring?
by
Noam David, Omry Sendik, Hagit Messer, and Pinhas Alpert
The potential of cellular network infrastructure
as a futuristic system for monitoring fog is introduced.
T
he Glossary of Meteorology (Glickman 2000)
defines fog as water droplets suspended in the
atmosphere near Earth’s surface that reduce visibility to less than 1 km. The intensity of fog can be
characterized by its liquid water content (LWC) and
its droplet number concentration ND or by the visibility
existing in the area observed during the occurrence
of the phenomenon. According to the World Meteorological Organization (WMO) visibility is defined
as the greatest distance in a given direction at which
a prominent black object can be seen and identified
against the sky at the horizon in daylight, or the greatest distance it could be seen and recognized at night
if the general illumination were raised to the level of
AFFILIATIONS: David
and A lpert—Department of Geosciences, Tel Aviv University, Tel Aviv, Israel; Sendik and Messer—
School of Electrical Engineering, Tel Aviv University, Tel Aviv,
Israel
CORRESPONDING AUTHOR: Dr. Noam David, Dept. of
Geosciences, Tel Aviv University, P.O. Box 39040, Tel Aviv
6997801, Israel
E-mail: [email protected]
The abstract for this article can be found in this issue, following the
table of contents.
DOI:10.1175/BAMS-D-13-00292.1
In final form 3 December 2014
©2015 American Meteorological Society
AMERICAN METEOROLOGICAL SOCIETY
normal daylight (WMO 2008). Visibility is related to
both the LWC and droplet number in a given volume
of air. In warm fog conditions (T > 0°C), and a given
LWC, an increase in ND results in decreased visibility.
An increase in LWC results in decreased visibility
as well. Additionally, prior field research found that
LWC increases with increasing N D (Gultepe et al.
2009). This being the case, visibility parameterizations for warm fog conditions should include both
LWC and ND (Meyer et al. 1980; Gultepe and Isaac
2004; Gultepe et al. 2006). Details on particle spectra,
chemical composition, definitions, and visibility issues
in warm and cold fog conditions can be found in the
literature (Klemm et al. 2005; Herckes et al. 2007;
Gultepe et al. 2009, 2014). Severe visibility limitations
associated with the phenomenon of fog may result in
acute transportation accidents, substantial property
losses, and human casualties. The total economic
damage caused by the impact of fog on the different
transportation modes in air, at sea, and on land can
be reasonably compared with that of winter storms
(Gultepe et al. 2009). Thus, reliable fog monitoring facilities are of extreme importance. Nevertheless, current observation techniques may suffer from obstacles
in providing a sufficient response to coping with this
hazard on a worldwide scale. Currently, predominant
monitoring means include satellites, visibility sensors,
transmissometers, and human-based observations.
Satellite systems are most advantageous due to their
OCTOBER 2015
| 1687
high spatial coverage. Their observations, however,
may suffer from difficulties in measuring near-surface
levels of fog, for example, as a result of middle- or
high-altitude clouds that obscure the ground-level
fog from the satellite’s vantage point (Gultepe et al.
2007a). Visibility sensors provide estimates based
on the principle of forward scattering of visual light.
These instruments, being relatively small in dimension, only acquire small, localized air samples and
may, therefore, provide observations that may not be
representative of the total areal fog patch. Thus, the
point measurements obtained by these instruments
should be used cautiously for larger area coverage
(Gultepe et al. 2007b). Transmissometers, composed
of a transmitting and a receiving unit, typically located
several tens of meters apart from each other, measure
both the scattering and the absorption coefficients of
light and can therefore provide accurate observations.
However, their implementation and maintenance
costs are extremely high. Consequently, these tools are
characteristically deployed only in specific locations
of interest such as airports. Human observers can
identify fog and estimate the visibility based on the appearance and obstruction of objects that are located at
known distances. However, this way of observation is
not only “not automatic” but also subjective as a result
of its psychophysical nature, as one human’s eye may
assess the visibility differently than another’s.
On the other hand, David et al. (2013b) have
recently shown that commercial microwave links
(MLs), which form the infrastructure of cellular communication networks, can provide crucial information concerning the appearance of dense fog and its
intensity. Cellular network infrastructure is widely
spread over the terrain at ground-level altitudes. As
these commercial networks operate at frequencies
of tens of gigahertz, they are highly sensitive to the
effects of weather phenomena and can be considered
as built-in facilities for atmospheric monitoring, as
was first demonstrated for rain-rate observations by
Messer et al. (2006) and Leijnse et al. (2007). Since
that point, rainfall monitoring using commercial
MLs has been extensively studied (Zinevich et al.
2008, 2009, 2010; Chwala et al. 2012; Wang et al.
2012; Doumounia et al. 2014). Notably, the received
signal level (RSL) measurements of these networks
are routinely logged by many of the communication
providers and can therefore be utilized for measuring
fog. The costs of using the presented technology are
particularly low, as opposed to complete installation
and positioning of specialized dedicated instruments, since the cellular network infrastructure is
already deployed in the field. At present, commercial
1688 |
OCTOBER 2015
microwave systems typically operate at frequencies
between around 6 and 40 GHz. As such, the already
existing links can, for the most part, observe only relatively heavy fog when its intensity (LWC, ND) is high
enough to be sensed by the given system. Modern
mobile radio networks have an increasing need for
high data rates and wide bandwidth; however, in
most current networks, capacity expansion through
coding and higher bandwidth have been maximized,
and no further increase can be achieved through
these means. To overcome this, higher frequencies
are being implemented as a solution for network access expansion. Radio links in bands as high as, for
example, 81–86 GHz are already being implemented
to fulfill these increased bandwidth requirements [for
further reading, see Csurgai-Horváth and Bitó (2010)
and Harsryd and Edstam (2011)]. These frequency
bands are highly sensitive to the effects of fog and can
potentially provide, for the first time, the opportunity
to monitor typical fog intensities on a wide scale, at
high resolution and low costs.
This work presents a theoretical comparison of
the sensitivity of commercial microwave systems in
detecting fog, when operating at the typical current
operating frequencies, 20 and 38 GHz, versus the 80GHz range, a frequency range that is being increasingly deployed. Real-data measurements of commercial
MLs operating in the 38-GHz band are also presented
to demonstrate the potential applicability of the technique in reality. The aim of this paper is, therefore, to
examine and quantify the expected improvement in
the potential of cellular network infrastructure as an
opportunistic system to monitor fog.
SCIENTIFIC BACKGROUND. The attenuation
effects of fog on microwave channels (dB km−1) are
well known and can be described by the following
relation:
,
(1)
where γF is the fog-induced attenuation, a function
of the LWC (g m−3) and the expression κ (dB km−1)
(g m−3)−1, which is dependent on the temperature T (K)
and the link frequency f (GHz). [For further reading
concerning the detailed functions comprising κ, refer
to International Telecommunication Union (2013).]
Figure 1, which is based on Eq. (1), describes the
attenuation as a function of microwave frequencies
induced by fog at typical liquid water contents, ranging between 0.01 and 0.4 g m−3 (Gultepe et al. 2007b).
Notably, the fog-induced attenuation increases with
increasing frequency.
Fig. 1. Transmission loss per 1 km at frequencies of up to 100 GHz due to fog at typical LWCs
[Eq. (1); International Telecommunication Union (2013)]. The horizontal dashed line, along
the 0.1-dB level, signifies a typical magnitude resolution (digital quantization error) of commercial MLs and specifically the one used during the current study. The yellow bars indicate
traditional microwave frequency ranges commonly used today, while blue denotes available
frequency bands being implemented more commonly worldwide in order to fulfill increased
bandwidth needs.
Previous studies, including, for instance, Kunkel
(1984) and Klemm et al. (2005), have shown a high
correlation between the LWC of fog and visibility.
Thus, given the fog LWC, an estimation of the visibility
can be derived (and vice versa). Additional studies
have shown the dependency of visibility (in addition
to its dependency on the LWC) on the microphysical
characteristics of the fog, particularly on the droplet
number per cubic volume ND (cm−3). The following
parameterization scheme, developed by Gultepe et al.
(2006) for warm fog conditions (T > 0°C), can be used
in order to obtain a rough visibility estimation (Vis;
km) given LWC and ND:
.(2)
The uncertainty in estimating visibility using this
parameterization scheme [Eq. (2)] was found to be
approximately 29% under the assumption that in the
case of visibility, fractional uncertainty is the sum of
AMERICAN METEOROLOGICAL SOCIETY
the fractional uncertainties in LWC and the parameter
ND, which in this case were 15% and 30%, respectively.
Based on the RSL acquired by the microwave
system, γF can be derived and, thus, using Eq. (1), the
LWC of the fog. As a result, given ND, a rough estimation of the visibility can be obtained using Eq. (2).
Preferably, ND can be measured directly using specialized equipment or estimated based on previous observations conducted, for example. This parameter may
also be obtained as a function of temperature when
saturation occurs but nucleation processes cannot be
ignored (Gultepe and Isaac 2004).
In this work the LWC levels were calculated using
theoretical simulations and the value of the parameter ND taken from the literature and substituted as a
constant in Eq. (2) (Gultepe and Isaac 2004; Gultepe
et al. 2006, 2007b). Consequently, the visibility estimations were acquired without testing their accuracy
directly (a task that is beyond the scope of this paper).
Thus, the visibility estimate derived is used only as an
order-of-magnitude guide.
OCTOBER 2015
| 1689
TWO-DIMENSIONAL SIMULATION TOOL:
CALCULATING THE MINIMUM DETECTAB LE LWC U S I NG M E AS U R E M E NTS
FROM MULTIPLE MLS. A simulation study has
been carried out in order to quantify the minimum
detectable LWC that can be sensed using a given set
of existing MLs deployed across Israel by one cellular
provider. The algorithm employed generates a twodimensional map of the minimum detectable LWC
based on the network’s given characteristics, that is,
the link locations, lengths, frequencies, and their RSL
quantization resolution.
To determine the effective detection threshold for
each link, it is assumed that the minimal fog-induced
attenuation can be sensed whenever the signal loss
caused is equal to the quantization interval ∆q (dB),
divided by the length of the intersection between
the fog patch and the link L (km). The minimum
detectable liquid water content, LWCMIN (g m−3), can
therefore be calculated as follows:
.(3)
Notably, at a given frequency and temperature, the
longer L is, the longer the distance that ∆q is divided
over, and as a result the effective sensitivity for unit
of distance is better, under the assumption made
here that the fog is homogenous along L. Surface
conditions—the properties and current state of the
soil, vegetation, or urban layout in the case of fog
over land, and the state of the surface of the sea for
marine fog—strongly influence the formation of fog,
its evolution, and its spatial pattern. The effect of
the surface on the fog can be direct, such as in cases
where it influences wind profiles, local circulations,
temperature, or humidity, or it may be indirect, such
as in cases when it modifies the radiative properties
of the atmosphere through microphysical processes
(Gultepe et al. 2007b). Here, the spatial shape of the
fog patch, its rotation angle and dimension, can be
modeled according to the user’s definition. It was
chosen to model the fog patch as a generalized ellipse of an area of about 15 km2 (e.g., Pagowski et al.
2004), which was inserted into the test site with its
major axis oriented north–south. The fog’s modeled
patch was swept across the area of the whole map we
wished to create. Then, for each point on the map,
the intersection between the link and the fog was
derived. Clearly, only the part of the link path that
actually intersects with the fog assists in sensing it. If
the length of a given intersection between a link and
a fog profile was greater than zero, then Eq. (3) was
1690 |
OCTOBER 2015
used to determine the minimum detectable LWC. In
cases when a group of MLs deployed over the same
location was covered by the same fog patch, the algorithm chose the longest link within the group (i.e.,
the most sensitive link per unit of distance) in order
to derive the minimum detectable value of LWC for
that specific domain.
SIMULATION RESULTS. The threshold fog LWC
values calculated using the algorithm described above
were compared for three different cases. The first
simulation was run using a real, already existing, set of
MLs operating in a frequency range of around 38 GHz.
During the second simulation study, the performance
of the same set of links was tested while a simulated
frequency of 20 GHz was chosen for the whole set of
links. In the third simulation the algorithm was run
using a simulated frequency of 80 GHz, illustrating
a future network designed to fulfill the increasing
demands of network access expansion. During each
of the three simulations the algorithm was applied
on a set of 696 links, deployed over an area of about
14,000 km2, with a quantization of 0.1 dB each, as
illustrated in Fig. 2a. The areal temperature was
taken to be 10°C. The sensitivity maps in Figs. 2b–d
present the minimum detectable LWCs, which were
derived by applying the same algorithm on the set of
links depicted in Fig. 2a at a simulated frequency of
20 GHz (Fig. 2b), at the original existing frequency
around 38 GHz (Fig. 2c) and a simulated frequency
of 80 GHz (Fig. 2d).
Under the same link deployment (illustrated in Fig.
2a), Fig. 3 presents the spatial coverage of fog that can
be obtained using each of the three frequency ranges
discussed here, at LWCs of 0.01, 0.05, and 0.1 (g m−3).
In accordance with the increase in the ability to
sense lighter fog, as depicted in the sensitivity maps
(Figs. 2b–d), the number of links capable of detecting the phenomenon is getting larger with increasing
frequency. As theoretically expected (International
Telecommunication Union 2013) given a certain LWC
and temperature, the sensitivity of the simulated 80GHz links improves on the sensitivity of the 38- and
20-GHz links by a factor of approximately 3.5 and 10,
accordingly. As a result, the spatial coverage that can
be achieved increases, as shown in Fig. 3. Specifically,
it can be seen from Fig. 3 that fog detection at the 80GHz frequency range was good for all three LWC values tested, while at the 38-GHz frequency detection
was good at LWC values of 0.1 and 0.05 (g m−3), and
in the 20-GHz case, fog detection is hardly possible.
Rough visibility estimates were derived using
Eq. (2) (Gultepe et al. 2006, 2009) based on the
Fig. 2. Simulation results: sensitivity. (a) A set of 696 links deployed by one cellular provider in Israel (updated to
Mar 2014). The MLs are deployed at altitudes between 5 and 200 m above the surface with lengths ranging from
40 m up to 4.1 km. One may notice the high variation in link density across the country. The urban area of the
city of Tel Aviv (central Israel) is characterized by a dense distribution of links as compared to the rural part of
the country (e.g., southern Israel), which has a less dense distribution. (b) The minimal LWC values that can be
detected by the given set of MLs as calculated given a simulated frequency of 20 GHz. The colored shadows illustrate the LWC threshold values according to the scale to the right of the figure. In several cases, the sensitivity
threshold values that were calculated for this frequency range (20 GHz) exceeded the range of physical values for
fog LWC and are indicated in this figure with the dark red–brown color that appears at the top region of the LWC
scale. (c) The minimal LWC values that can be detected by the microwave network, calculated given a frequency
range of around 38 GHz, which is the actual link frequency of the set as deployed by the cellular provider. (d) The
minimal LWC values that can be detected as calculated using a simulated frequency of 80 GHz.
minimum detectable LWC values that were calculated according to the sensitivity of the system at
each geographic location (as presented in Figs. 2b–d).
These estimates denote a rough threshold value of
visibility below which the microwave network can
sense fog for each one of the three frequency ranges
demonstrated here. For this calculation we used the
fixed value of ND = 150 cm−3 (Gultepe and Isaac 2004;
Gultepe et al. 2006, 2009). Two visibility threshold
values were derived for each of the given three operating frequencies. The first is an upper visibility value
that was calculated based on the system of the longest
links deployed in a certain region, where the system
is the most sensitive per unit of distance [according
to Eq. (3)]. The second threshold value (which is
lower) derived for the same operating frequency was
calculated for a different area where the link system
is the least sensitive per unit of distance (because in
that particular area, the links are of shorter lengths).
The visibility threshold estimates derived were found
to be in the range from 750 (in the case of the most
sensitive MLs) down to 100 m (in the case of the less
sensitive ones), at 80 GHz. At 38 GHz, the rough visibility threshold estimates derived were 500 and 50 m,
respectively. Of the 20-GHz links, the most sensitive
ones were capable of sensing only relatively heavy fog,
where the estimated visibility is 150–200 m and less.
AMERICAN METEOROLOGICAL SOCIETY
Since at this frequency range (20 GHz) the system is
relatively insensitive to the effects of fog, in several
cases, the minimum detectable LWC values that were
calculated exceeded the range of physical values for
fog (see Fig. 2b).
REAL-DATA SPATIAL MEASUREMENTS OF
FOG FROM COMMERCIAL MICROWAVE
NETWORKS. This section presents real-data measurements, taken in the field, from tens of commercial
MLs, operating in the 38-GHz band, during heavy
fog, demonstrating the potential applicability of the
technology. The demonstration uses data from an
event, presented in David et al. (2013b), that took place
from the late evening hours of 9 December 2005 to the
morning hours of the following day. The extreme fog
event caused major disruptions to schedules at the Ben
Gurion International Airport in Israel, as well as flight
cancellations, because of the poor visibility. Figure 4a
presents an Meteosat Second Generation (MSG) image
produced by the Clouds–Aerosols–Precipitation Satellite Analysis Tool (CAPSAT; Lensky and Rosenfeld
2008) for the night of the event, around 0130 UTC,
the time we will focus on here. The image shows the
massive spatial dimensions of the fog, indicated by a
white shadow, covering hundreds of square kilometers
from the Sinai Peninsula in Egypt to the southern and
OCTOBER 2015
| 1691
central coastal regions of Israel. Figures 4b and 4c
show the deployment map of the microwave system,
and the locations of the different ground stations,
drawn over zoomed-in portions of the satellite image
covering the Tel Aviv and Jerusalem areas, respectively. In particular, it can be seen that an area of low
stratus cloud/fog was detected by the satellite system
around Tel Aviv, as indicated by the whitened pixels
(Fig. 4b), whereas the phenomenon was not detected
in the vicinity of Jerusalem
at that time (Fig. 4c). Two
human observers located
in Beit Dagan (35 m MSL)
and at the Ben Gurion airport (41 m MSL), as well as
three transmissometers at
the airport, also detected
fog between the hours of
2200 UTC 9 December and
0700 UTC 10 December
and limited visibility that
dropped to as low as hundreds, and even tens, of
meters, and particularly
when the satellite image was
taken [for further details,
see David et al. (2013b)].
On the other hand, a human observer located in
Jerusalem (815 m MSL) did
not detect fog and reported
visibility in the range of
10–15 km between the same
hours of the event.
Fig. 3. Simulation results: spatial detection. Spatial detection capability of the
system for identifying fog at three different LWCs of 0.01, 0.05, and 0.1 (g m−3)
using frequency ranges of (a)–(c) 20, (d)–(f) 38, and (g)–(i) 80 GHz. Each panel
represents the whole study area. Each column of the matrix represents a constant LWC value, as shown above [(a)–(c)].The respective visibility estimates
are indicated beneath the LWC values and were calculated using Eq. (2) for ND =
150 cm−3 (Gultepe et al. 2006, 2009) and T = 10°C.The Mediterranean Sea is colored blue and the white shadow illustrates the area where fog was detected by
the system during each simulation test. Black domains signify regions where fog
was not identified. Notably, the 80-GHz links introduce high-spatial-detection
capability, at various fog LWC levels.
1692 |
OCTOBER 2015
The microwave system
in the observed areas. Let
us focus, then, on the Tel
Aviv area, where fog was
identified by the specialized instruments, and the
Jerusalem region, where
no fog was detected.
Commercial MLs are deployed in both of these regions, and the system from
which measurements were
taken is built of links ranging between 100 m and 4 km
in length, and operating in
the 38-GHz band. Each link
provides one momentary
RSL measurement once every 24 h around 0130 UTC.
The microwave system in
the Tel Aviv area (Fig. 4b, in
Fig. 4. Regional satellite image and locations of the specialized measuring instruments and MLs.
(a) MSG image of the observed region taken at 0127 UTC and presented as a “night microphysical” color scheme using CAPSAT (Lensky and Rosenfeld 2008). The fog is indicated as a white
shadow. (b) A magnified satellite image of the Tel Aviv area with a map of the commercial
MLs deployed in the region (green lines) and the locations of the different ground stations. (c)
A microwave link map (black lines) overlaid on a magnified satellite image of the Jerusalem
area. The whitened pixels in (b) indicate detection of fog/low stratus cloud in that area by the
satellite system as opposed to the Jerusalem area, where the phenomenon was not detected.
green) has 88 links deployed over 46 paths, while the
Jerusalem system (Fig. 4c, in black) consists of 36 links
over 19 paths (multiple links can operate over the same
path by using separate frequencies or polarizations).
Results. Figure 5 shows the attenuation measurements from the links as a function of their length D
(km) as taken in the areas around Tel Aviv (Fig. 5,
left, in green) and Jerusalem (Fig. 5, right, in black).
Each point in the figures indicates a single measurement received by each of the links around 0130 UTC
10 December 2005.
To calculate the attenuation on each link, a baseline
reference signal level has to be set and attenuation
derived with respect to it. For each link, this reference point was chosen to be the median of the RSL
measurements from the link during that month.
The equations in each panel in Fig. 5 describe the
calculated linear approximations. The slope of the
measurement graph from the foggy region, as well as
AMERICAN METEOROLOGICAL SOCIETY
the y-axis intercept, can be seen to be larger than the
equivalent values for the nonfoggy region (where both
parameters tend to zero approximately). The differences in slope are the result of attenuation increasing
with link length in the case where the microwave
signals interacted with the foggy medium (Tel Aviv
area), while in the area where the phenomenon did
not exist (Jerusalem), attenuation did not increase
with link length. During fog, relative humidity (RH)
is exceptionally high, and as a result, a thin layer of
water can condense directly on the microwave antennas, increasing the attenuation measured by the
system. The y-axis intercept of the graphs represents
here an imaginary infinitesimal distance between
the microwave antennas. The value calculated for
this parameter based on the system measurements
in the foggy area indicates that this wet-antenna
phenomenon occurred (Henning and Stanton 1996;
Harel et al. 2015). This being the case, calculating
these two parameters (slope, y-axis intercept) by
OCTOBER 2015
| 1693
Fig. 5. ML observations. Attenuation measurements as a function of link length taken by the microwave systems
deployed in the areas around (left) Tel Aviv and (right) Jerusalem. According to the measurements of the meteorological stations in the observed areas, no precipitation (rain, hail, etc.) that may have affected microwave
system measurements fell during the event. Between 0120 and 0140 UTC, the RH was found to be in the range
of 97%–100% in the Tel Aviv area and 33% in the Jerusalem area. The Pearson correlation coefficient between
the attenuation measurements and the propagation pathlengths in Tel Aviv area (left) was found to be r = 0.6 with
a p value < 0.01 (Neter et al. 1996, 640–645). System magnitude resolution is 0.1 dB.
using system measurements allows for the detection
of cases of fog, as shown in David et al. (2013b), while
also using meteorological side information, including
humidity gauges for detecting conditions of high relative humidity (RH > 95%) and offsetting its effects on
the system (International Telecommunication Union
2005) and rain gauges or other techniques to rule
out different forms of precipitation (rain, hail, etc.)
that also affect received signal levels in the network.
Additionally, the ability to derive a rough estimation
for visibility from the slopes of the measurement
graphs in the fog-affected area was shown [using Eqs.
(1) and (2), given the areal temperature]. In the case
presented here, it was found to be in the range between
30 and 70 m and matched the measurements of human
observers and the specialized equipment in the area in
order of magnitude. For further reading describing the
method in detail, the uncertainty calculations, further
examples of microwave measurements, etc., the reader
is kindly referred to David et al. (2013b).
The microwave system that provided the real measurements shown here stores data once every 24 h. It is
important to note that there are systems that routinely
store measurements in higher temporal resolutions,
for example, every 1 or every 15 min (e.g., Rayitsfeld
et al. 2012; Liberman et al. 2014). The measurements
presented here are from the microwave data that were
available from this fog event.
In this section, two specific subregions were
examined using real data, out of the entire area
investigated theoretically through simulation (in
"Two-dimensional simulation tool: Calculating the
minimum detectable LWC using measurements from
multiple MLs" and "Simulation results"). A future challenge is, therefore, to further investigate the potential
1694 |
OCTOBER 2015
applicability in all areas where the network is deployed,
over wide regions, for a large number of events, while
comparing the results to ground-truth measurements,
with the goal of testing the ability to spatially map the
phenomena using current and future frequency bands.
DISCUSSION. At frequencies of tens of gigahertz,
rainfall has a predominant effect on the microwave
channel compared to those induced by other hydrometeors such as water vapor and fog. Thus, commercial MLs have been shown to be particularly useful
for rainfall monitoring (Rayitsfeld et al. 2012; David
et al. 2013a; Sendik and Messer 2015). A preliminary
study, by David et al. (2013b), showed the feasibility
of monitoring dense fog (with LWCs in the range
0.5–0.8 g m−3) using multiple commercial MLs operating around the 38-GHz frequency range. The real
measurements presented here also demonstrate the
potential for identifying or ruling out the occurrence
of the phenomena in space, in different regions.
The results of the simulation point to the potential
for great improvement in the capabilities of future microwave systems to detect lighter fog, including cases
in the typical LWC range for fog (LWC < 0.4 g m−3;
Gultepe et al. 2007b, 2009)], as a result of theoretically
expected attenuation effects when moving to higher
frequencies.
Certain previous research has shown that the LWC
can vary widely with altitude and in some cases grow
by up to an order of magnitude above the values that
exist near the surface (Pinnick et al. 1978; Kunkel 1984;
Fuzzi et al. 1992). However, further research on this
topic is required (Klein and Dabas 2014). On the other
hand, commercial microwave systems are deployed at
elevations of between several meters and up to tens or
hundreds of meters above ground level. In practice,
microwave units are installed on masts, buildings, towers, and in some cases on hills and mountains (Chwala
et al. 2012; David et al. 2013a). Thus, the system can
provide measurements from a range of elevations.
Measurement effectiveness can suffer, though, in
cases where the microwave propagation path is, for
example, above the fog layer, particularly since this
phenomenon is often created in low-lying areas. The
transition to higher frequencies provides a solution
for network access expansion, a need that is greatly
increasing, especially in densely populated areas. On
the other hand, since the attenuation caused by rainfall
increases with the operating frequency, this transition
generally goes hand in hand with a move to shorter
links, which are more resilient to precipitation. As a
result of these factors, the use of relatively short links
operating at high frequencies is especially suited to urban areas. This is in contrast to rural areas, where the
transmission distances required between base stations
are longer and therefore the tendency is to use lower
frequencies. The potential for fog monitoring using
commercial MLs is thus expected to be distributed
across a certain area, according to these constraints.
In the current paper the analysis of system performance in the simulation was a theoretical one,
using an existing, given set of commercial MLs. It
is important to note that other interference factors
exist and may lead to inexact measurements (see, e.g.,
Zinevich et al. 2010). This can be seen, for the realdata case presented here as well, in the scatter of the
points representing the microwave measurements
around the linear fit in Fig. 5. These factors include
the existence of other atmospheric phenomena in
parallel with the fog, such as precipitation that affects the RSL more dominantly or condensation of a
thin layer of water on the outside of the microwave
transmission units, causing excess signal attenuation. Additionally, changes to the baseline reference level can be caused, for example, by changes
in humidity (David et al. 2009; Chwala et al. 2014),
and add to the uncertainty in the measurements.
Therefore, in addition to the RSL measurements,
David et al. (2013b) used supplemental side information (humidity and rain gauge data) to identify the
fog. In addition to these environmental sources of
perturbation, technical factors affecting the ability
to acquire accurate measurements exist, including
white noise, the magnitude resolution of the MLs,
and additional characteristics of the system. For
example, in some commercial microwave systems
transmission power is not constant and can vary
(increase or decrease). In these cases this parameter
AMERICAN METEOROLOGICAL SOCIETY
needs to be taken into account in addition to the RSL
to estimate the baseline reference level. Some other
systems with a coarser magnitude resolution than
the one used here (0.1 dB) exist. For example, some
systems have a magnitude resolution of 1 dB. In the
course of the expected transition to higher-frequency
links, the sensitivity to fog will increase, and the
ability to contend with these types of interference
will improve as well. These topics, which lie outside
of the scope of the current paper, are left for future
research. That being said, in communication systems
that are densely deployed, the links measuring the
phenomenon are varied in pathlength and operating
frequencies. The amount of data provided simultaneously is immense and can partially compensate for
the measurement quality of each link on its own. The
diversity of the link system being used as the sensor array for monitoring allows us to overcome the
different interference factors. Thus, for example, by
using a large number of short and long links, which
are simultaneously measuring the fog in the same
area, the attenuation as a result of condensation on
the microwave units themselves can be separated
from the attenuation caused by the fog itself, utilizing the principle that the shorter the link, the more
dominant the effect of condensation on the antenna,
and vice versa. This is done, as is presented here, for
example, by calculating a parametric fit.
Additional research could also focus on the combination of meteorological side information from
ground-based sensors or satellites with the microwave
link measurements. For instance, RSL from MLs could
be used to check whether fog in satellite images is,
in fact, present close to the surface (or, perhaps, is
stratus cloud found at higher altitudes). Naturally,
this depends on the availability of these dedicated
sensor data. Furthermore, commercial MLs were
found to be efficient in their ability to differentiate
between cases of rainfall and cases where no rain existed (e.g., Rayitsfeld et al. 2012), as well as between
different types of precipitation [pure rain and sleet;
Cherkassky et al. (2014)], through the use of signal
processing techniques. Additionally, the feasibility of
using commercial microwave systems for measuring
atmospheric water vapor utilizing frequencies around
the absorption line of this phenomenon at 22.235 GHz
was shown (David et al. 2009, 2011). It is possible, then,
in future research, to develop and investigate the use of
these methods and others to minimize or completely
remove the need for side information for fog detection.
In the simulation carried out here, homogeneity of the
fog patches introduced into the area where the links are
deployed was assumed. In reality, fog intensity changes
OCTOBER 2015
| 1695
in space (e.g., David et al. 2013b). Given a suitably sensitive link network (with respect to the fog affecting it),
a future challenge would be to investigate the ability to
create a two-dimensional map of the distribution of fog
intensity in space using real RSL measurements—that
is, in addition to a simple detection of whether the
phenomenon occurred in a given area, in a similar
fashion to rainfall-intensity mapping, which is already
carried out using this technology (e.g., Zinevich et al.
2008, 2009; Goldshtein et al. 2009; Overeem et al.
2013; Liberman et al. 2014). Additionally, the possibility of improving the precision of the model used
here (International Telecommunication Union 2013)
to estimate fog intensity by adding information about
the microphysical characteristics of the phenomenon
(e.g., ND or fog droplet size distribution) can be also
included in the scope of future research.
CONCLUSIONS. The goal of the work presented
here is to reveal the future potential that exists in commercial microwave systems, where higher frequencies,
which are more sensitive to fog, are starting to be used.
This being the case, we carried out a theoretical simulation to model the planned future state. The novelty of
the current paper is, foremost, conceptual, presenting
the window of opportunity being opened by the planned
future engineering design, for monitoring fog with high
resolution. Additionally, unique algorithms were developed, as part of this research, allowing and theoretically
demonstrating for the first time the following:
1)the creation of a 2D map of the sensitivity threshold to fog in a region, given an arbitrary link deployment, and
2)2D detection (there was or was not fog) in an area
of interest by using commercial microwave links.
In summar y, t he fol low ing points ca n be
highlighted:
• When cellular network infrastructure is shifted to
operate in higher-frequency bands, its use for fog
monitoring can be more reliable.
• The method suggested here can potentially be
used for real-time fog prediction.
• Fog prediction is a complex process and strongly
depends on model physical parameterizations
(Gultepe et al. 2007b). This suggests that future
improvements exist for the method proposed
here, as detailed in the discussion section.
• Validation of the proposed technology needs to
be done using well-planned field projects that use
state-of-the-art measurements.
1696 |
OCTOBER 2015
Fog is an important atmospheric parameter for
varied fields of interest, from meteorology to transportation safety to its effects on economics and more.
The method proposed here for fog monitoring using
cellular network infrastructure can potentially provide information of extraordinary value.
ACKNOWLEDGMENTS. The authors are grateful to
Yariv Dagan, Idit Alexandrovitz, Yaniv Koriat, Ido Inbar,
Shahar Shilyan, Eli Levi, and Yosi Eisenberg (Cellcom);
Nissim Dvela (Pelephone); and Haim Ben Shabat, Amir
Shor, Haim Mushvilli, and Yoav Bar Asher (Orange) for
providing the microwave data, pro bono, for our research.
We deeply thank our research team members (Tel Aviv
University): Yonatan Ostrometzky, Ori Auslender, Rana
Samuels, Roi Reich, Yaakov Cantor, Itay Zlotnik, Ronen
Radian, Lior Gazit, and Elad Heiman for their fruitful
cooperation and discussions.
Our sincere acknowledgments to the Israeli Meteorological
Service and the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) for providing meteorological and satellite data A special thank you
to the weather specialist Asaf Rayitsfeld from the Israeli
Meteorological Service for the practical consultation and
advice.
REFERENCES
Cherkassky, D., J. Ostrometzky, and H. Messer, 2014:
Precipitation classification using measurements
from commercial microwave links. IEEE Trans.
Geosci. Remote Sens., 52, 2350–2356, doi:10.1109
/TGRS.2013.2259832.
Chwala, C., and Coauthors, 2012: Precipitation observation using microwave backhaul links in the
alpine and pre-alpine region of southern Germany.
Hydrol. Earth Syst. Sci., 16, 2647–2661, doi:10.5194
/hess-16-2647-2012.
—, P. Kunstmann, S. Hipp, and U. Siart, 2014: A
monostatic microwave transmission experiment
for line integrated precipitation and humidity remote sensing. Atmos. Res., 144, 57–72, doi:10.1016/j
.atmosres.2013.05.014.
Csurgai-Horváth, L., and J. Bitó, 2010: Fog attenuation
on V band terrestrial radio and a low-cost measurement setup. Future Network and Mobile Summit
2010 Conference Proceedings, P. Cunningham
and M. Cunningham, Eds., IIMC International
Information Management Corporation, 47.
David, N., P. Alpert, and H. Messer, 2009: Novel method
for water vapour monitoring using wireless communication networks measurements. Atmos. Chem. Phys.,
9, 2413–2418, doi:10.5194/acp-9-2413-2009.
—, —, and —, 2011: Humidity measurements using commercial microwave links. Advanced Trends in
Wireless Communications, M. Khatib, Ed., In-Tech
Open Access Publisher, 65–78.
—, —, and —, 2013a: The potential of cellular
network infrastructures for sudden rainfall monitoring in dry climate regions. Atmos. Res., 131, 13–21,
doi:10.1016/j.atmosres.2013.01.004.
—, —, and —, 2013b: The potential of commercial
microwave networks to monitor dense fog- feasibility
study. J. Geophys. Res. Atmos., 118, 11 750–11 761,
doi:10.1002/2013JD020346.
Doumounia, A., M. Gosset, F. Cazenave, M. Kacou, and F.
Zougmore, 2014: Rainfall monitoring based on microwave links from cellular telecommunication networks:
First results from a West African test bed. Geophys.
Res. Lett., 41, 6016–6022, doi:10.1002/2014GL060724.
Fuzzi, S., and Coauthors, 1992: The Po Valley Fog
Experiment 1989: An overview. TeIlus, 44B, 448–468,
doi:10.1034/j.1600-0889.1992.t01-4-00002.x.
Glickman, T., Ed., 2000: Glossary of Meteorology. 2nd
ed. Amer. Meteor. Soc., 855 pp.
Goldshtein, O., H. Messer, and A. Zinevich, 2009:
Rain rate estimation using measurements from
commercial telecommunications links. IEEE
Trans. Signal Process., 57, 1616–1625, doi:10.1109
/TSP.2009.2012554.
Gultepe, I., and G. A. Isaac, 2004: Aircraft observations
of cloud droplet number concentration: Implications
for climate studies. Quart. J. Roy. Meteor. Soc., 130,
2377–2390, doi:10.1256/qj.03.120.
—, M. D. Müller, and Z. Boybeyi, 2006: A new
visibility parameterization for warm-fog applications in numerical weather prediction models. J.
Appl. Meteor. Climatol., 45, 1469–1480, doi:10.1175
/JAM2423.1.
—, M. Pawgoski, and J. Reid, 2007a: A satellite-based fog
detection scheme using screen air temperature Wea.
Forecasting, 22, 444–456, doi:10.1175/WAF1011.1.
—, and Coauthors, 2007b: Fog research: A review
of past achievements and future perspectives.
Pure Appl. Geophys., 164, 1121–1159, doi:10.1007
/s00024-007-0211-x.
—, and Coauthors, 2009: The fog remote sensing and
modeling field project. Bull. Amer. Meteor. Soc., 90,
341–359, doi:10.1175/2008BAMS2354.1.
—, and Coauthors, 2014: Ice fog in Arctic during
FRAM–Ice Fog Project: Aviation and nowcasting
applications. Bull. Amer. Meteor. Soc., 95, 211–226,
doi:10.1175/BAMS-D-11-00071.1.
Harel, O., N. David, P. Alpert, and H. Messer, 2015: The
potential of microwave communication networks to
detect dew using the GLRT—Experimental study.
AMERICAN METEOROLOGICAL SOCIETY
IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.,
doi:10.1109/JSTARS.2015.2465909, in press.
Harsryd, J., and J. Edstam, 2011: Microwave capacity
evaluation. Ericsson Review, Stockholm, Sweden.
[Available online at www.ericsson.com/res/docs
/review/Microwave-Capacity-Evolution.pdf.]
Henning, R. E., and J. R. Stanton, 1996: Effects of dew
on millimeter-wave propagation. Proceedings of the
IEEE Southeastcon '96: Bringing Together Education,
Science and Technology, IEEE, 684–687, doi:10.1109/
SECON.1996.510158.
Herckes, P., H. Chang, T. Lee, and J. L. Collet Jr.,
2007: Air pollution processing by radiation fogs.
Water Air Soil Pollut., 181, 65–75, doi:10.1007
/s11270-006-9276-x.
International Telecommunication Union, 2005:
Attenuation by atmospheric gases. Rec. ITU-R
P.676-6. 23 pp. [Available online at www.itu.int/rec
/R-REC-P.676-6-200503-S/en.]
—, 2013: Attenuation due to clouds and fog. Rec. ITU-R
P.840-6, 4 pp. [Available online at www.itu.int/rec
/R-REC-P.840-6-201309-I/en.]
Klein, C., and A. Dabas, 2014: Relationship between
optical extinction and liquid water content in
fogs. Atmos. Meas. Tech., 7, 1277–1287, doi:10.5194
/amt-7-1277-2014.
Klemm, O., T. Wrzesinsky, and C. Scheer, 2005: Fog
water f lux at a canopy top: Direct measurement
versus one-dimensional model. Atmos. Environ.,
39, 5375–5386, doi:10.1016/j.atmosenv.2005.05.041.
Kunkel, B., 1984: Parameterization of droplet terminal velocity and extinction coefficient in fog
models. J. Appl. Meteor. Climatol., 23, 34–41,
doi:10.1175/1520-0450(1984)023<0034:PODTVA
>2.0.CO;2.
Leijnse, H., R. Uijlenhoet, and J. Stricker, 2007: Rainfall
measurement using radio links from cellular communication networks. Water Resour. Res., 43,
W03201, doi:10.1029/2006WR005631.
Lensky, I. M., and D. Rosenfeld, 2008: Clouds-AerosolsPrecipitation Satellite Analysis Tool (CAPSAT).
Atmos. Chem. Phys., 8, 6739–6753, doi:10.5194
/acp-8-6739-2008.
Liberman, Y., R. Samuels, P. Alpert, and H. Messer,
2014: New algorithm for integration between
wireless microwave sensor network and radar for
improved rainfall measurement and mapping.
Atmos. Meas. Tech., 7, 3549–3563, doi:10.5194
/amt-7-3549-2014.
Messer, H., A. Zi nev ich, a nd P. A lper t, 20 06:
Environmental monitoring by wireless communication networks. Science, 312, 713, doi:10.1126/
science.1120034.
OCTOBER 2015
| 1697
Meyer, M. B., J. E. Jiusto, and G. G. Lala, 1980:
Measurements of visual range and radiation-fog
(haze) microphysics. J. Atmos. Sci., 37, 622–629,
doi:10.1175/1520-0469(1980)037<0622:MOVRAR
>2.0.CO;2.
Neter, J., M. H. Kutner, C. Nachtsheim, and W.
Wasserman, 1996: Applied Linear Statistical Models.
4th ed. McGraw Hill, 720 pp.
Overeem, A., H. Leijnse, and R. Uijlenhoet, 2013:
Country-wide rainfall maps from cellular communication networks. Proc. Natl. Acad. Sci. USA, 110,
2741–2745, doi:10.1073/pnas.1217961110.
Pagowski, M., I. Gultepe, and P. King, 2004: Analysis
and modeling of an extremely dense fog event
in southern Ontario. J. Appl. Meteor., 43, 3–16,
doi:10.1175/1520-0450(2004)043<0003:AAMOAE
>2.0.CO;2.
Pinnick, R. G., D. L. Hoihjelle, G. Fernandez, E. B. Stenmark,
J. D. Lindberg, and G. B. Hoidale, 1978: Vertical structure in atmospheric fog and haze and its effects on visible and infrared extinction. J. Atmos. Sci., 35, 2020–
2032, doi:10.1175/1520-0469(1978)035<2020:VSIAF
A>2.0.CO;2.
Rayitsfeld, A., R. Samuels, A. Zinevich, U. Hadar, and
P. Alpert, 2012: Comparison of two methodologies
for long term rainfall monitoring using a commercial
microwave communication system. Atmos. Res., 104–
105, 119–127, doi:10.1016/j.atmosres.2011.08.011.
1698 |
OCTOBER 2015
Sendik, O., and H. Messer, 2015: A new approach to
precipitation monitoring: A critical survey of existing technologies and challenges. IEEE Signal Process.
Mag., 32, 110–122, doi:10.1109/MSP.2014.2309705.
Wang, Z., M. Schleiss, J. Jaffrain, A. Berne, and J.
Rieckermann, 2012: Using Markov switching models
to infer dry and rainy periods from telecommunication microwave link signals. Atmos. Meas. Tech., 5,
1847–1859, doi:10.5194/amt-5-1847-2012.
World Meteorological Organization, 2008: Guide
to Meteorological Instruments and Methods of
Observation. 7th ed. WMO-8. [Available online at
www.wmo.int/pages/prog/gcos/documents/gruan
manuals/CIMO/CIMO_Guide-7th_Edition-2008.pdf.]
Zinevich, A., P. Alpert, and H. Messer, 2008: Estimation
of rainfall fields using commercial microwave
communication networks of variable density.
Adv. Water Resour., 31, 1470–1480, doi:10.1016/j
|.advwatres.2008.03.003.
—, H. Messer, and P. Alpert, 2009: Frontal rainfall
observation by a commercial microwave communication network. J. Appl. Meteor. Climatol., 48,
1317–1334, doi:10.1175/2008JAMC2014.1.
—, —, and —, 2010: Prediction of rainfall intensity measurement errors using commercial microwave communication links. Atmos. Meas. Tech., 3,
1385–1402, doi:10.5194/amt-3-1385-2010.